Reply to Gille et al.

to the editor: We thank Dr. Gille and collaborators for their comment (1) to our review (2). The following statements in their letter have to be discussed: 1) In contrast to the findings of Vogel et al. (3), the in vivo oxygen dissociation curve is not left-shifted in the patients of their recent study (4). 2) There is no marked increase in methemoglobin concentration in their patients as suggested by us for Vogel’s results. 3) Oxygen saturation is calculated as the sum of oxyhemoglobin (O₂Hb) and carboxyhemoglobin (COHb).

IN VIVO OXYGEN DISSOCIATION CURVE

Four topics need to be considered, as follows.

Number of Measurements

The number of measurements is much smaller in Gille’s than in Vogel’s investigation [19,463 from 43 patients with COVID-19 and 828 critically ill patients with acute respiratory failure (3)]. Gille et al. (4) evaluated 474 samples from 100 patients with COVID-19 and 100 subjects with other respiratory problems. In addition to the affinity difference between the patient groups, a significant decrease of P₅₀ relative to the standard value calculated according to Severinghaus (5, 6) was found in the subjects infected with SARS-CoV-2 by Vogel et al. (3).

Severity of Illness

A marked difference between the two studies is notable in terms of hemoglobin concentration: although the mean concentration is within the low normal range (median and interquartile ranges 14.0 [12.6–15.2] g/dL) in the investigation of Gille et al. (4), the patients in the article by Vogel et al. (3) are anemic (COVID-19: 8.1 ± 1.2 SD g/dL; comparison group: 9.4 ± 2.0 SD g/dL).

Methods

Gille et al. (4) excluded values with saturations above 97% from P₅₀ calculations. Indeed, the variation in $\text{P}{\text{O}}_2$ is large while $\text{S}{\text{O}}_2$ changes very little in this part of the curve. As also the Bohr effect disappears at high saturation [described, e.g., by Garby et al. (7) and Meier et al. (8)], this approach is in any case reasonable. Severinghaus (6) even suggests using only samples between 20% and 80% $\text{S}{\text{O}}_2$ for the P₅₀ calculation. Vogel et al. (3) compare immense numbers of measurements ranging between 20% and 100% $\text{S}{\text{O}}_2$ with similar mean values of ∼94% $\text{S}{\text{O}}_2$ for COVID-19 and their control group of critically ill patients. Therefore, the outliers in their study are present in both groups and the detected difference should be real. Additionally, pH deviation from 7.4 and its variation are very small (7.382 ± 0.077 SD), therefore possibly false Bohr effect corrections are negligible.

Results

There is indeed no difference in standardized P₅₀ between patients with COVID-19 and the negative control group of patients who do not have COVID in the overall statistical evaluation presented in Table 1 in Gille’s first paper (4). But when considering Fig. 2 (mean P₅₀ over 18 days in the hospital), there appears a tendency for lower mean values in the patients with COVID-19, especially between days 3 and 15 as compared with the control group (∼2 mmHg difference on days 8–10). The differences are not significant, but this does not prove that they do not exist. It is well possible that only the variability is too large or the number of measurements was too low. In addition, differences in the non-COVID-19 control groups (critically ill patients with acute respiratory failure, in the case of Vogel et al., vs. a mixed population of patients with infections, airway disease, interstitial lung disease, heart failure, and surgical interventions, etc., in Gille’s study) and the varying use of medians and interquartile ranges (Table 1) or mean values ± standard errors (Fig. 1) in contrast to only mean values in the article of Vogel et al. (3) makes the comparison somewhat difficult. In view of the statistical difficulties to prove absence (rather than presence) of an effect, the conclusion in the title of Gille’s first paper (1), “The affinity of hemoglobin for oxygen is not altered during COVID-19,” appears somewhat overstated, in contrast to the question mark punctuating the title of their Letter to the Editor.

METHEMOGLOBIN

The findings of Gille et al. (4) together with those of Vogel et al. (9) communicated in their letter to the editor indeed exclude MetHb as main factor for a left shift of the in vivo oxygen dissociation curve in COVID-19. We previously already suggested a possible influence of varying nitric oxide (NO)-binding to Hb as an alternative explanation for this finding (10). Interestingly, erythroblasts possess angiotensin-converting enzyme 2 (ACE2) receptors and are therefore potentially vulnerable to COVID-19 infection in cases of viremia (11).

CALCULATION OF OXYGEN SATURATION AND P₅₀

With respect to methods, it seems puzzling that Gille et al. (4) added COHb to O₂Hb for calculation of P₅₀: “s is the combined O₂ and CO saturation.” This approach will likely increase the calculated O₂ saturation in both patient groups in their study and therefore reduce P₅₀. The authors state that this was done according to Dash et al. (12), yet we were unable to find the corresponding formula in that paper. To our knowledge, this approach is also not used in the calculations by Severinghaus or other authors. Severinghaus’s curve (5, 13) is based on volumetric measurements of oxygen in the samples (therefore the COHb content was unknown) or on photometric measurements at the specific absorption wavelength for O₂Hb.

COHb is omitted for the calculation of P₅₀ in many investigations according to Kirk et al. (14). They describe that the inclusion of COHb in the saturation term is used in Instrumentation Lab apparatus, but not in American Optical Company oximeters. Vogel et al. calculated P₅₀ only for $\text{S}{\text{O}}_2$ (personal communication).

As the affinity for carbon monoxide (CO) is 200 times higher than for O₂ (15) and influences oxygen affinity, its inclusion in the calculation of P₅₀ appears rather unorthodox. Fortunately, the content of COHb is small in the study of Gille et al. (4) and therefore will not affect their P₅₀ calculation very much.

CONCLUSIONS

The controversial findings in the studies by Vogel and Gille and the resulting discussion highlight the need for further definite studies on this topic. In particular, additional measurements of the concentrations of key modulators of oxygen affinity in red cells such as [H⁺] and 2,3-bisphosphoglycerate (2,3-BPG) are required to understand the mechanisms behind a potential shift of the oxygen dissociation curve in patients with COVID-19.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

D.B. drafted manuscript; D.B., W.B., and W.M.K. edited and revised manuscript; D.B., W.B., and W.M.K. approved final version of manuscript.